The so-called buried point refers to the collection of some information in specific processes in the application, which is used to track application usage, and subsequently used to further optimize products or provide data support for operations, including Visits, Visitors, Time On Site), Page Views (also known as page browsing) and Bounce Rate (also known as bounce rate). Such information collection can be roughly divided into two types: track this virtual page view and track this button by an event.
1. key indicators
我们先看看无论是 APP 还是 H5 都会关注的指标，了解这些指标的计算方法的细微差异以及复杂性，换个角度来思考埋点的意义。【源自：精通 Web Analytics 2.0】
1.1 Visits and visitors
Visits and visitors are indicators that almost all applications need to be counted, and they are also the most basic indicators.
对于应用的统计来说，希望统计的是访客（Vistors）。访问（Visits）是指会话层，用户打开应用花一段时间浏览又离开，从指标定义来说这杯称之为一个会话（Session）。一次会话（Session 或 Visit）是打开应用的第一个请求（打开应用）和最后一个请求决定的。如果用户打开应用然后放下手机或是离开电脑，并在接下来 30 分钟内没有任何动作，此次会话自动结束，算作一次访问或会话期。
在计算访客时，埋点上报的数据是尽可能接近真实访客的人数。对于独立访客这个指标，这里还是需要强调一下，独立访客数并不是真实独立的人，因此收集数据时必须知道独立访客虽然能够很好的反映使用应用的真实访问者的数量，但不等于使用应用的真实人数。
The reason is that repeated installation of apps or modification of mobile phone parameters will affect the indicators of independent visitors. Independent visitors rely on cookies. When a user opens the application, the application will create an independent Cookie on the person's terminal. The Cookie will be retained, but it will inevitably be manually cleaned by the user or the Cookie will be disabled, resulting in inconsistent use of the application Cookie by the same user. Therefore, independent visitors can only be highly close to the real number of people using the application.
1.2 stay time
The dwell time is used to measure the time a user spends on a certain page or visit (session) of the application.
Page stay time represents the time spent on each page; for example, the home page is the time from entering the home page (10:00) to leaving the home page to enter the next page (10:01), and the home page stay time is calculated as 1 minute. Page A is 2 minutes. The entry time of page B (10:03) and the departure time is not recorded. At this time, the calculation is 0. The handling of this special situation requires special attention at the burial point. Again, don't try to collect absolutely accurate data. Learn to use incomplete data and use it flexibly.
The stay time of the application represents the time spent in one visit (session), which is calculated as the visit time of all pages. In the same process, the stay time of the application is 4 minutes.
1.3 bounce rate
There are still many methods to calculate bounce rate in various companies. The most commonly used one is the proportion of sessions visited by a single page. This scenario means that a user comes and visits a page and then leaves. Think about the user's mental picture when using it: Open the application, think about something, then close the application or even uninstall it. How scary this scene is, which is why the bounce rate indicator is so concerned.
The bounce rate can be broken down into two levels: one is the bounce rate of the entire application, the other is the bounce rate of key landing pages, and even the bounce rate of search keywords. The indicator of bounce rate is very operable. By counting the bounce rate, you can directly discover problems on the page and find problems with keywords.
1.4 exit rate
The exit rate is specific to pages. The goal of this indicator is very simple: how many users leave a certain page, and the main users report how users leave the application. The fastest way to discover which pages need to be improved. (In some processes, when completing the standard process, the one with the highest exit rate is reflected in the positive direction on the last page of the standard process. Don't think that a high exit rate is a bad thing)
1.5 conversion rate
Isn't we investing so much in products just to measure output? So for e-commerce applications, is there any indicator that deserves more attention than conversion rate? Conversion rate is calculated by dividing a certain output by unique visitors or visits. For e-commerce products, it is the number of users submitting orders divided by unique visitors.
The calculation of conversion rates may seem simple, but it is the closest data collection to the business among the buried sites. This is also the indicator that best reflects the embedding skills, and a calculation method needs to be formulated based on business characteristics. The number of submitted orders/number of visitors is the most basic conversion rate. The conversion rate can also be divided into levels and specify the user path, such as: the number of submitted orders/number of visitors completing a certain path.
** Try to find a path and think about how the conversion rate data is obtained and what kind of data is collected at the burial site? **
1.6 participation
Participation is not an indicator, but a series of indicators. Visit depth and visit frequency are all indicators to measure participation. The reason why participation is included as an indicator is that we hope that everyone will understand that the combination of indicators will subsequently produce chemical reactions and discover the nature of the real thing.
2. Ways to bury data points
There are two mainstream ways to bury sites now:
The first type: my company develops and injects code statistics into the product and sets up corresponding background queries.
The second type: third-party statistical tools, such as Youmeng, Baidu Mobile, Rubik's Cube, App Annie, talking data, etc.
If your data comes from the second type, then the tool you use should also be a third-party statistical tool. There are no data products in the future, so use these products well. Let's talk about the first method of embedding points here. How to bury data needs to be designed according to the task flow and product goals of your own product.
2.1 Front end buried point
代码埋点出现的时间很早了，在 Google Analytics 年代，就已经出现了类似的方案了。目前，国内的主要第三方数据分析服务商，如百度统计、友盟、TalkingData 等都提供 iOS、Android、Web 等主流平台的代码埋点方案。原理就是在 APP 或者界面初始化的时候，初始化数据分析的 SDK，然后在某个事件发生时就调用 SDK 里面相应的数据发送接口发送数据。现在业界有吹嘘无埋点的其实并不是没有埋点，而是不需要手动埋点，其实是从接入 SDK，数据就一直都在收集。有兴趣读一读提供的 SDK，会更了解前端的埋点，收集的信息。包括现在也有了不断的演化统计埋点的那些事。
2.2 Back-end buried point
Back-end buried points are also server-end buried points. In addition to recording the log of the interface, some parameters are added to the interface to transfer the information layer by layer. Because the transformation that relies on the interface is usually used to supplement the statistics that the front-end buried points cannot achieve, most of my company's data statistics are based on the coexistence of front and back buried points.
3. Contents of the burial site
看完关键的这些指标后，有没有发现埋点的来源也大致分为两部分，一部分是统计应用页面访问情况，即页面统计；另外一部分是统计应用内的操作行为，及自定义事件统计。
数据产生就是在每次页面浏览或是点击，滑动等事件发生时都上报一条数据，包括页面信息，控件信息，设备信息，用户信息等，为了将用户行为串联，需要确保有一个全局唯一的ID串联访问的顺序。
Subsequent statistics can serially count the user's behavior flow to produce the desired result.
4. Notes on data from buried sites
- Don't pursue perfection too much
One thing about buried point data is crucial. Burying points is to make better use of the data. Don't try to get accurate data. You want to get high-quality buried point data. The bounce rate discussed above is an example. Get what you can get. Data, use imperfect data to achieve the next step, and pursue high quality rather than accuracy. This is a place where many data products can easily fall into the trap, so you should always remind yourself.